摘要
注塑成型中,把熔前速度恒定(熔体均匀填充模具)作为设定注射速度参数的重要依据。针对熔前速度无法测量问题,提出对于一类注塑制品,用模腔压力随时间线性增加来近似熔前速度恒定。其次,利用实验数据,设计并建立了模腔压力递归神经网络(RNN)模型。在此基础上,提出一种注射速度设定曲线的分级优化策略,其中曲线分级是利用Douglas-Peucker方法自动确定速度分级数目和位置,优化过程则采取基于滑动窗口的迭代优化方法来实现。实验结果表明,该优化方法可以根据模腔复杂程度自动选择合适的注射曲线分级数目,并且其优化精度较常规等间隔分级优化方法更高。
Uniform melt front （MFV） velocity throughout the filling of the mold cavity is recommended as the design criterion for injection profile.Due to the difficulty of measuring MFV,a criterion that cavity pressure followed a ramp with time was proposed as an approximate substitute to constant MFV in this study.Then a recurrent neural network was developed for estimating the cavity pressure from online measurable variables,based on which the profile problem was transformed to an optimization problem for the predicted cavity pressure to track a given ramp.Subsequently the velocity profile was treated as a piecewise ramp one with respect to the increment of screw stroke,and its turning points and the slopes of each section were the optimization variables.The Douglas-Peucker method was introduced to determine the number and positions of separation points of the profile and an iterative optimization method with sliding window was presented to solve the optimization problem.Experimental results showed that this optimization method could automatically give the number of pieces of injection velocity profile and its precision of optimization was higher than the regular optimization method with equal fixed intervals.
出处
《化工学报》
EI
CAS
CSCD
北大核心
2010年第5期1163-1168,共6页
CIESC Journal
基金
国家重点基础研究发展计划项目(2009CB320601)
国家自然科学基金项目(60774068)~~
